Gait Recognition via Machine Learning
Aybüke KEÇECİ Armağan YILDIRAK Kaan ÖZYAZICI Gülşen AYLUÇTARHAN Onur AĞBULUT İbrahim ZİNCİR
AbstractThe basis of biometric authentication is that each person's physical and behavioral characteristics can be accurately defined. Many authentication techniques were developed for years. Human gait recognition is one of these techniques. This article was studied on HugaDB database which is a human gait data collection for analysis and activity recognition (2017, Chereshnev and Kertesz-Farkas). Combined activity data of different people were collected in HugaDB database (2017, Chereshnev and Kertesz-Farkas). The activities are walking, running, sitting and standing (2017, Chereshnev and KerteszFarkas). The data were collected with devices such as wearable accelerometer and gyroscope (2017, Chereshnev and KerteszFarkas). Only the walking dataset of the HugaDB was used artificial neural network-based method for real-time gait analysis with the minimal number of Inertial Measurement Units (2018, Sun et al). In this paper, each person is considered as a different class because there are multiple users' gait data in the database and some machine learning algorithms have been applied to walking, running, standing and sitting data. The best algorithms are chosen from the algorithms applied to the HugaDB data and the results are shared.